name: git-commit description: Create professional git commits following conventional commit format with proper staging and verification group: git keywords: [commit, git, conventional commits, staging, version control, commits] version: 1.0.0 author: docent

Git Commit Runbook

This runbook guides you through creating professional, well-structured git commits that follow project conventions and represent logical units of work.

Purpose

Create git commits that:

Follow conventional commit format
Represent atomic, logical units of work
Match existing project style and conventions
Pass pre-commit hooks
Are independently reviewable and revertible

Prerequisites

Git repository initialized
Working directory with changes to commit
Write access to current branch

Procedure

Step 1: Examine Current State

Action: Gather complete context about changes and repository state

git status                    # See untracked files and modifications
git diff                     # Review unstaged changes
git diff --staged           # Review already staged changes
git log --oneline -10       # Check recent commit style
git branch --show-current   # Confirm current branch

Analysis Required:

Before proceeding, analyze:

What files have changed and why
Whether changes represent ONE logical unit or MULTIPLE distinct units
What conventional commit type applies (feat, fix, docs, etc.)
Whether any files should be excluded (temporary files, secrets, build artifacts)
If multiple units exist, how to group them logically

Step 2: Determine Commit Strategy

Single Commit When:

All changes serve the same purpose
Changes are interdependent and cannot function separately
Small, focused change set
Emergency hotfix requiring atomic deployment

Multiple Commits When:

Changes span different conventional commit types (feat + fix + docs)
Independent features/fixes that could be reviewed separately
Configuration changes mixed with code changes
Substantial documentation alongside code changes
Changes affect different modules/components independently
Refactoring mixed with new functionality
Test additions separate from implementation

Step 3: Analyze Existing Commit Style

Action: Identify project conventions

Look for patterns in git log:

Does repo use conventional commits? (feat:, fix:, chore:, etc.)
What's typical message length and format?
Are scopes used? (feat(auth):, fix(api):)
Are there ticket/issue references required?
Body format and detail level?

Match the existing style exactly.

Step 4: Stage Changes

For Single Commit:

# Review and stage relevant files
git add src/feature.ts
git add tests/feature.test.ts
git add README.md

# Or stage all tracked changes
git add -u

# Or stage specific hunks interactively
git add -p src/complex-file.ts

For Multiple Commits:

Stage files in logical groups:

# Example: Stage only configuration files first
git add config/ .env.example docker-compose.yml

Exclusion Checklist:

Do NOT stage:

Temporary files (.tmp,.log, *.swp)
Build artifacts (dist/, build/, node_modules/)
OS files (.DS_Store, Thumbs.db)
Editor files (.vscode/, .idea/)
Debug code or commented-out code
Secrets or API keys

Step 5: Craft Commit Message

Format:

type(scope): brief description in imperative mood

[optional body explaining WHY, not HOW]

[optional footer with breaking changes or issue refs]

Conventional Commit Types:

feat: - New feature for users
fix: - Bug fix
docs: - Documentation changes only
style: - Formatting, whitespace, no code change
refactor: - Code restructuring without behavior change
test: - Adding or updating tests
chore: - Maintenance, dependencies, tooling
perf: - Performance improvements
ci: - CI/CD configuration changes
build: - Build system or external dependency changes

Quality Standards:

Subject line: 50 chars max, imperative mood ("add" not "added")
Body: 72 chars per line, explain motivation and context
Focus: Business impact and "why", not technical "how"

Examples:

# Simple feature
git commit -m "feat: add user profile page"

# Feature with scope
git commit -m "feat(auth): implement JWT token refresh"

# Bug fix with body
git commit -m "fix: prevent race condition in file upload

The previous implementation didn't properly lock the upload queue,
causing concurrent uploads to overwrite each other. Added mutex
to ensure sequential processing."

# Breaking change
git commit -m "feat!: migrate to v2 API endpoints

BREAKING CHANGE: All API endpoints now require /v2/ prefix"

Step 6: Create Commit

git commit -m "feat: add user authentication system

- Implement JWT token generation
- Add login/logout endpoints
- Create user session management"

For Multi-line Messages:

Use heredoc for clean formatting:

git commit -m "$(cat <<'EOF'
feat: implement full-text search

Add search functionality across all content types:
- Product catalog search with fuzzy matching
- User directory search
- Document full-text indexing

Improves user experience by providing unified search interface.
EOF
)"

Step 7: Verify Commit Success

Action: Confirm commit was created correctly

git log -1 --oneline          # Confirm commit exists
git show --stat HEAD          # Show committed files
git status                    # Should show clean working directory

Success Criteria:

✅ Commit appears in git log with correct message
✅ All intended files included
✅ No unintended files included
✅ Working directory clean (or only intentionally uncommitted files remain)
✅ Commit message follows project conventions

Step 8: Handle Pre-commit Hook Failures (If Applicable)

If hooks modify files:

git status                          # See what hook changed
git diff                           # Review hook modifications
git add -u                         # Stage hook changes
git commit --reuse-message=HEAD    # Retry with same message

If hooks fail with errors:

Read error output carefully
Common issues: linting failures, test failures, formatting
Fix the root cause
Retry commit after fixes

Do NOT use --no-verify unless explicitly necessary.

Multi-Commit Workflow

When analysis reveals multiple logical units:

Step 1: Identify Logical Boundaries

Group related files by:

Purpose/feature
Commit type (feat, fix, docs, test, etc.)
Dependency order (earlier commits should not break functionality)

Step 2: Create Commits Sequentially

Example: Feature with Tests and Documentation

# Commit 1: Core implementation
git add src/features/search/ src/utils/search-helpers.ts
git commit -m "feat: implement full-text search functionality"
git log -1 --oneline  # Verify

# Commit 2: Tests
git add tests/search/ tests/fixtures/search-data.json
git commit -m "test: add search feature test coverage"
git log -1 --oneline  # Verify

# Commit 3: Documentation
git add README.md docs/search.md
git commit -m "docs: document search API and configuration"
git log -1 --oneline  # Verify

Example: Bug Fix with Refactoring

# Fix first (higher priority)
git add src/validator.ts
git commit -m "fix: correct email validation regex"

# Then refactor
git add src/validator.ts src/types.ts
git commit -m "refactor: extract validation helpers"

Example: Configuration and Feature

# Infrastructure first (feature may depend on it)
git add package.json package-lock.json
git commit -m "chore: add lodash dependency"

# Then feature
git add src/utils/collection-helpers.ts
git commit -m "feat: add collection utility functions"

Step 3: Use Partial Staging for Complex Files

When single file contains multiple logical changes:

git add -p src/app.ts  # Interactively stage hunks

# Stage only auth-related changes
git commit -m "feat: add authentication middleware"

# Stage remaining changes
git add src/app.ts
git commit -m "refactor: reorganize route registration"

Step 4: Verify Each Commit

Run verification protocol after EACH commit in multi-commit workflow.

Pre-Commit Verification Checklist

Before committing, verify:

✅ No temporary files (.tmp,.log, .DS_Store, etc.)
✅ No debug code (console.log, debugger, print statements)
✅ No hardcoded secrets or API keys
✅ No sensitive data or personal information
✅ Commit message follows project style
✅ Changes represent logical, atomic unit
✅ All staged files are intentional

Error Recovery

Undo Last Commit (Keep Changes):

git reset --soft HEAD~1

Amend Last Commit Message:

git commit --amend -m "corrected message"

Unstage Specific File:

git reset HEAD <file>

Undo All Staging:

git reset HEAD

Branch Considerations

Default to main branch unless specified
In feature branch, ensure commit aligns with branch purpose
For worktrees, verify correct branch before committing
Check branch with git branch --show-current

Common Patterns

Documentation Update

git add README.md docs/
git commit -m "docs: update installation instructions"

Dependency Update

git add package.json package-lock.json
git commit -m "chore: update dependencies to latest versions"

Configuration Change

git add .eslintrc.json prettier.config.js
git commit -m "chore: update linting rules"

Performance Improvement

git add src/api/users.ts
git commit -m "perf: optimize user query with database indexing

Reduced query time from 450ms to 23ms by adding composite index
on (email, status) columns."

Breaking Change

git commit -m "feat!: migrate to ESM module format

BREAKING CHANGE: All imports now require .js extension.
Update import statements from:
  import {foo} from './bar'
to:
  import {foo} from './bar.js'
"

Validation

After completing commits:

✅ git log shows commit(s) with proper messages
✅ git status shows clean working directory or only intentionally uncommitted files
✅ Each commit represents an atomic, logical unit
✅ Commit messages follow project conventions
✅ Pre-commit hooks passed (if applicable)

Notes

Focus on business value and user impact in commit messages
Explain "why" in body, not "how" (code shows "how")
Keep commits atomic - each should be independently functional
Batch related changes in same commit, separate unrelated changes
When in doubt, prefer multiple focused commits over one large mixed commit

git-commit

SKILL.md